[Quantization] [Performance] Enable Marlin GEMM kernels for the calibration-free RTN-based quantization (vllm-project#26051)

Signed-off-by: Alex Kogan <alex.kogan@oracle.com>
Signed-off-by: Alex Kogan <82225080+sakogan@users.noreply.github.com>
Signed-off-by: 0xrushi <6279035+0xrushi@users.noreply.github.com>

Author: sakogan

vllm/model_executor/layers/quantization/rtn.py

@@ -6,21 +6,16 @@
 from collections.abc import Callable
 from typing import Any, Optional
 
+import numpy as np
 import torch
-import torch.nn.functional as F
 from torch.nn.parameter import Parameter
 
 from vllm.logger import init_logger
-from vllm.model_executor.layers.fused_moe import (
-    FusedMoE,
-    FusedMoEConfig,
-    FusedMoEMethodBase,
-)
 from vllm.model_executor.layers.fused_moe.config import (
+    FusedMoEConfig,
     FusedMoEQuantConfig,
-    int4_w4a16_moe_quant_config,
-    int8_w8a16_moe_quant_config,
 )
+from vllm.model_executor.layers.fused_moe.layer import FusedMoE, FusedMoEMethodBase
 from vllm.model_executor.layers.linear import (
     LinearBase,
     LinearMethodBase,
@@ -31,6 +26,12 @@
     QuantizationConfig,
     QuantizeMethodBase,
 )
+from vllm.model_executor.layers.quantization.utils import replace_parameter
+from vllm.model_executor.layers.quantization.utils.marlin_utils import (
+    apply_rtn_marlin_linear,
+    marlin_make_workspace_new,
+)
+from vllm.scalar_type import scalar_types
 
 logger = init_logger(__name__)
 """By default, use 8 bit as target precision, but it can be 
@@ -41,6 +42,9 @@
 overridden by setting the RTN_GROUP_SIZE envvar
 """
 GROUP_SIZE = os.getenv("RTN_GROUP_SIZE", "128")
+"""Global Marlin workspace shared by all modules
+"""
+workspace = None
 
 
 class RTNConfig(QuantizationConfig):
@@ -60,6 +64,10 @@ def __init__(
                 f"supported for RTN, but got {self.weight_bits} bits."
             )
 
+        self.quant_type = (
+            scalar_types.uint8b128 if self.weight_bits == 8 else scalar_types.uint4b8
+        )
+
     def __repr__(self) -> str:
         return (
             f"RTNConfig(weight_bits={self.weight_bits}, group_size={self.group_size})"
@@ -221,24 +229,32 @@ def create_weights(
         layer.output_size_per_partition = output_size_per_partition
 
     def process_weights_after_loading(self, layer: torch.nn.Module) -> None:
-        fix_weights(layer, "weight")
+        """Repack weights and scales for Marlin kernels."""
+        weight_bits = self.quant_config.weight_bits
+
+        weight, scale = repack_weights(layer.weight, layer.scale, weight_bits)
+
+        replace_parameter(layer, "weight", weight)
+        replace_parameter(layer, "scale", scale)
+
+        init_workspace(layer.weight.device)
 
     def apply(
         self,
         layer: torch.nn.Module,
         x: torch.Tensor,
         bias: torch.Tensor | None = None,
     ) -> torch.Tensor:
-        qweight = layer.weight
-        scale = layer.scale
-
-        weight = rtn_dequantize(qweight, scale)
-        out = F.linear(x, weight)
-        del weight
-        if bias is not None:
-            out.add_(bias)
-
-        return out
+        return apply_rtn_marlin_linear(
+            input=x,
+            weight=layer.weight,
+            weight_scale=layer.scale,
+            workspace=workspace,
+            quant_type=self.quant_config.quant_type,
+            output_size_per_partition=layer.output_size_per_partition,
+            input_size_per_partition=layer.input_size_per_partition,
+            bias=bias,
+        )
 
 
 class RTNMoEMethod(FusedMoEMethodBase):
@@ -315,28 +331,27 @@ def create_weights(
         set_weight_attrs(w2_weight, extra_weight_attrs)
 
     def process_weights_after_loading(self, layer: torch.nn.Module) -> None:
+        """Repack weights and scales for Marlin kernels."""
         weight_bits = self.quant_config.weight_bits
-        fix_weights(layer, "w13_weight", weight_bits == 4)
-        fix_weights(layer, "w2_weight", weight_bits == 4)
+
+        w13_weight, w13_scale = repack_weights(
+            layer.w13_weight, layer.w13_scale, weight_bits
+        )
+        replace_parameter(layer, "w13_weight", w13_weight)
+        replace_parameter(layer, "w13_scale", w13_scale)
+
+        w2_weight, w2_scale = repack_weights(
+            layer.w2_weight, layer.w2_scale, weight_bits
+        )
+        replace_parameter(layer, "w2_weight", w2_weight)
+        replace_parameter(layer, "w2_scale", w2_scale)
+
+        init_workspace(layer.w13_weight.device)
 
     def get_fused_moe_quant_config(
         self, layer: torch.nn.Module
     ) -> FusedMoEQuantConfig | None:
-        weight_bits = self.quant_config.weight_bits
-        group_size = self.quant_config.group_size
-        assert weight_bits == 4 or weight_bits == 8
-        config_builder = (
-            int4_w4a16_moe_quant_config
-            if weight_bits == 4
-            else int8_w8a16_moe_quant_config
-        )
-        return config_builder(
-            w1_scale=layer.w13_scale,
-            w2_scale=layer.w2_scale,
-            w1_zp=None,
-            w2_zp=None,
-            block_shape=[0, group_size],
-        )
+        return None
 
     def apply(
         self,
@@ -366,8 +381,6 @@ def apply(
         if enable_eplb:
             raise NotImplementedError("EPLB not supported for `RTNMoEMethod` yet.")
 
-        from vllm.model_executor.layers.fused_moe import fused_experts
-
         topk_weights, topk_ids, _ = FusedMoE.select_experts(
             hidden_states=x,
             router_logits=router_logits,
@@ -383,18 +396,22 @@ def apply(
             indices_type=self.topk_indices_dtype,
         )
 
-        return fused_experts(
+        return torch.ops.vllm.fused_marlin_moe(
             x,
             layer.w13_weight,
             layer.w2_weight,
-            topk_weights=topk_weights,
-            topk_ids=topk_ids,
-            inplace=True,
-            activation=activation,
+            getattr(layer, "w13_bias", None),
+            getattr(layer, "w2_bias", None),
+            layer.w13_scale,
+            layer.w2_scale,
+            router_logits,
+            topk_weights,
+            topk_ids,
+            quant_type_id=self.quant_config.quant_type.id,
             apply_router_weight_on_input=apply_router_weight_on_input,
             global_num_experts=global_num_experts,
             expert_map=expert_map,
-            quant_config=self.moe_quant_config,
+            workspace=workspace,
         )
 
 
@@ -504,18 +521,133 @@ def rtn_dequantize(tensor: torch.Tensor, scale: torch.Tensor) -> torch.Tensor:
     return input_deq
 
 
-def fix_weights(layer: torch.nn.Module, param_name: str, reshape: bool = False):
-    """torch.compile does not know how to deal with a Parameter subclass
-    (aka RTNParameter). As we don't really need RTNParameters for the
-    forward pass, we replace them with equivalent instances of Parameters.
+def _get_perms():
+    perm = []
+    for i in range(32):
+        perm1 = []
+        col = i // 4
+        for block in [0, 1]:
+            for row in [
+                2 * (i % 4),
+                2 * (i % 4) + 1,
+                2 * (i % 4 + 4),
+                2 * (i % 4 + 4) + 1,
+            ]:
+                perm1.append(16 * row + col + 8 * block)
+        for j in range(4):
+            perm.extend([p + 256 * j for p in perm1])
+
+    perm_arr = np.array(perm)
+    interleave = np.array([0, 2, 4, 6, 1, 3, 5, 7])
+    perm_arr = perm_arr.reshape((-1, 8))[:, interleave].ravel()
+    perm_tensor = torch.from_numpy(perm_arr)
+    scale_perm = []
+    for i in range(8):
+        scale_perm.extend([i + 8 * j for j in range(8)])
+    scale_perm_single = []
+    for i in range(4):
+        scale_perm_single.extend([2 * i + j for j in [0, 1, 8, 9, 16, 17, 24, 25]])
+    return perm_tensor, scale_perm, scale_perm_single
+
+
+_perm, _scale_perm, _scale_perm_single = _get_perms()
+
+
+def pack_for_marlin(weight, scale, qbits):
+    batch = weight.shape[0]
+
+    n = weight.size(1)
+    k = weight.size(2)
+    groupsize = k // scale.size(2)
+
+    tile = 16
+    s = scale.permute(0, 2, 1)  # transpose
+    w = weight.permute(0, 2, 1)  # transpose
+    if groupsize != k:
+        w = w.reshape((batch, -1, groupsize, n))
+        w = w.permute(0, 2, 1, 3)
+        w = w.reshape((batch, groupsize, -1))
+        s = s.reshape((batch, 1, -1))
+
+    if groupsize != k:
+        w = w.reshape((batch, groupsize, -1, n))
+        w = w.permute(0, 2, 1, 3)
+        w = w.reshape((batch, k, n)).contiguous()
+        s = s.reshape((batch, -1, len(_scale_perm)))[:, :, _scale_perm]
+    else:
+        s = s.reshape((batch, -1, len(_scale_perm_single)))[:, :, _scale_perm_single]
+    s = s.reshape((batch, -1, n)).contiguous()
+    w = w.reshape((batch, k // tile, tile, n // tile, tile))
+    w = w.permute((0, 1, 3, 2, 4))
+    w = w.reshape((batch, k // tile, n * tile))
+    res = w
+    res = res.reshape((batch, -1, _perm.numel()))[:, :, _perm].reshape(res.shape)
+    if qbits == 4:
+        q = torch.zeros(
+            (batch, res.shape[1], res.shape[2] // 2), dtype=torch.int8, device=w.device
+        )
+        for i in range(2):
+            q |= res[:, :, i::2] << 4 * i
+        q = q.reshape(batch, -1, n).contiguous()
+    else:
+        q = res.clone()
+        q[:, :, 2::8] = res[:, :, 4::8]
+        q[:, :, 3::8] = res[:, :, 5::8]
+        q[:, :, 4::8] = res[:, :, 2::8]
+        q[:, :, 5::8] = res[:, :, 3::8]
+        q = q.reshape(batch, -1, n).to(torch.int8).contiguous()
+
+    return q, s
+
+
+def repack_8bit_into_32bit(input):
+    output = torch.zeros(
+        (input.shape[0], input.shape[1], input.shape[2] // 4),
+        dtype=torch.int32,
+        device=input.device,
+    )
+    for i in range(4):
+        output |= (input[:, :, i::4] & 0xFF).to(torch.int32) << 8 * i
+
+    return output
+
+
+def repack_weights(qweight, scale, weight_bits):
+    batch_present = len(qweight.shape) == 3
+    if not batch_present:
+        qweight = qweight.unsqueeze(0)
+        scale = scale.unsqueeze(0)
+
+    if weight_bits == 4:
+        """Unpack two 4-bit values from each byte.
+        """
+        qweight_unpacked = torch.empty(
+            (qweight.shape[0], qweight.shape[1] * 2, qweight.shape[2]),
+            dtype=torch.uint8,
+            device=qweight.device,
+        )
+        for i in range(2):
+            qweight_unpacked[:, :, i::2] = ((qweight << 4 * (1 - i)) >> 4).reshape(
+                qweight.shape[0], qweight.shape[1] * 2, qweight.shape[2] // 2
+            )
+    else:
+        qweight_unpacked = qweight
+
+    qweight_packed, scale_packed = pack_for_marlin(qweight_unpacked, scale, weight_bits)
+    """Marlin kernels expect tensors in int32 format in a certain shape
     """
-    old_weight = getattr(layer, param_name)
-    assert isinstance(old_weight, RTNParameter)
-    data = old_weight.data.data
+    qweight_repacked = repack_8bit_into_32bit(qweight_packed.to(torch.uint8))
+    qweight_reshaped = qweight_repacked.reshape(
+        qweight.shape[0], qweight.shape[2] // 16, -1
+    )
+    if not batch_present:
+        qweight_reshaped = qweight_reshaped.squeeze(0)
+        scale_packed = scale_packed.squeeze(0)
+
+    return qweight_reshaped, scale_packed
 
-    delattr(layer, param_name)
 
-    if reshape:
-        data = data.reshape(old_weight.shape[0], old_weight.shape[1] * 2, -1)
-    new_weight = Parameter(data=data, requires_grad=False)
-    layer.register_parameter(param_name, new_weight)
+def init_workspace(device):
+    global workspace
+    if workspace is None:
+        workspace = marlin_make_workspace_new(device, 4)

vllm/model_executor/layers/quantization/utils/marlin_utils.py

@@ -528,3 +528,48 @@ def apply_awq_marlin_linear(
     )
 
     return output.reshape(out_shape)
+
+
+def apply_rtn_marlin_linear(
+    input: torch.Tensor,
+    weight: torch.Tensor,
+    weight_scale: torch.Tensor,
+    workspace: torch.Tensor,
+    quant_type: ScalarType,
+    output_size_per_partition: int,
+    input_size_per_partition: int,
+    bias: torch.Tensor | None = None,
+    use_fp32_reduce: bool = USE_FP32_REDUCE_DEFAULT,
+) -> torch.Tensor:
+    reshaped_x = input.reshape(-1, input.shape[-1])
+    out_shape = input.shape[:-1] + (output_size_per_partition,)
+
+    use_atomic_add = should_use_atomic_add_reduce(
+        m=reshaped_x.size(0),
+        n=output_size_per_partition,
+        k=reshaped_x.size(1),
+        device=input.device,
+        dtype=input.dtype,
+    )
+
+    output = ops.gptq_marlin_gemm(
+        reshaped_x,
+        None,
+        weight,
+        bias,
+        weight_scale,
+        None,
+        None,
+        None,
+        None,
+        workspace,
+        quant_type,
+        size_m=reshaped_x.shape[0],
+        size_n=output_size_per_partition,
+        size_k=input_size_per_partition,
+        use_atomic_add=use_atomic_add,
+        use_fp32_reduce=use_fp32_reduce,
+        is_zp_float=False,
+    )
+
+    return output.reshape(out_shape)